wafer.space Community - 🏗️-die-sorter

namibj

Wdym, "data" that's "useful to capture"?

sorry, started replying and seems I got distracted haha, it's all useful, isn't it

At the base level, it's always nice to have something to look back on if there is some kind of issue somewhere. Additionally, Tim had wanted some kind of data on what shot in what wafer a die may have come from, it's difficult if something ever happens in the middle and you end up off by one or something, but it's in there anyways.

namibj

How about some light setup that enhances the effective contrast? A simple polarization filter on the camera and one on the lamp with their angles and the angle of the light (between normal and parallel to the plane of the die) tweaked while looking at the camera feed probably does wonders. Also, consider throwing a tiny variant from a modern AlexNet/MobileNet evolution at the task using some classic binary/ternary classifier layer over it's embedding just trained with manual labels ("good placement"/"too close to edge"/"literally overlapping the edge") and fed with a suitably carefully handled cropping to keep it from looking at irrelevant environment.

I ordered some more lights and stuff, I'll give it a shot. I haven't worked with ML models yet, but again, that's one of the reasons to just keep all the data so you may have a set to train against. Thanks for the idea

namibj

What do you mean, "up camera"? What's it looking at?

Up camera is a fixed camera in a known location, we can move the robot above, knowing where the nozzle is, and correct pick errors. It's pretty common in a lot of pcb pnp machines.

Hey all, I was running the machine and was asked to post a little vid. I'll post more in the near future. I have a few I want to finish quickly, then do some small upgrades (better lighting and some small things) Currently picking some @BreakingTaps dies. Basic setup is

robot arm
cartesian bot below wafer that pushes die off tape
other misc boards that power solenoids, or lights, or motors and a game controller

All the boards are networked over canbus HUD is cpython bridged onto bus as well. All control is handled through a web interface (Starlette)

up camera finds die on nozzle
feeder_camera locates the pocket -- place happens between offsets from these
down camera is on arm, does mostly calibration and location of the wafer (the dicing house sticks them on there all willy nilly, it's bad) Like I said, more coming soon, feel free to ask questions

3

1

@Andrew Wingate are those run #2 wafers or testing on run #1 wafers?

tnt

@Andrew Wingate are those run #2 wafers or testing on run #1 wafers?

Run #1 Hoping to get Run #2 Wed hopefully, maybe, hopefully

5:35 a.m.

Dicing house has them, they wanted to back grind again, so it's taking longer than we hoped.

2

Visualizing some of the pick data Hopefully will be interesting to see improvements over time

1

@BreakingTaps - This might be interesting content for future videos?

Tim 'mithro' Ansell

@BreakingTaps - This might be interesting content for future videos?

I will have better stuff in the future. Posting stuff more often than more better

definitely! this is awesome stuff, love seeing all this machine dev and behind the scenes. absolutely bet other folks would want to see it in a video too

1

I'm just going to dump these here.

3

2

9:15 a.m.

My setup leaves a lot to be desired, sorry. Shots are patterned in a lawnmower pattern going up, over, then back down, over and repeat

these are beautiful!

2

9:19 a.m.

9:21 a.m.

9:22 a.m.

and the eyecandy still keeps coming! (edited)

Is this backlit ?

tnt

Is this backlit ?

No, they're front lit

1

9:29 a.m.

@Tholin I tried to give something decent. I'll try again later, I want to get the sort underway

1

I taped out these patterns that are just horizontal arrays of 1u wide metal strips, but with increasing spacing, to see if I could build a light wavelength filter. And you can see it sortof working.

9:38 a.m.

In both the video and photos, the left handside of each pattern is some shade of red, shifting blue in hue towards the other end.

Tholin

In both the video and photos, the left handside of each pattern is some shade of red, shifting blue in hue towards the other end.

Ah, gorgeous!!

9:40 a.m.

Not sure why video is not rendering for me, if others are having issues I will try to convert to h.264

One half of this block is metal strips in a consistent 1u spacing, and it consistently reflects mostly blue light, so I think this is working.

3

@Tholin more stills. Seems different things refract at different angles. But doesn't seem like discord is honoring the pic numbering. When I get to your die, I'll snatch one out and get some when I can hold it better by itself.

Yeah, that seems about right. The light direction most likely need to be aligned with the filter.

10:03 a.m.

This pattern of metal strips of increasing widths is also doing interesting things, but more subtle

2

Gee, and all my project does is boring electrical stuff.

4

Tim Edwards

Gee, and all my project does is boring electrical stuff.

They are all wonderful!

die shots!

1

@BreakingTaps - I think you should get together with @Tholin to show off how her work to more people and demonstrate that there is more fun and interesting things that can be done.

Tim Edwards

Gee, and all my project does is boring electrical stuff.

We need that SRAM and IO cells!

Dammit, @Tholin & @Andrew Wingate - I had things I was going to do and now I've somehow spent an hour gawking at pretty silicon

4

@Andrew Wingate I am forever grateful, these are splendid!

1

@Tim 'mithro' Ansell : On my very first full-chip design (1992) I had a very pad-limited chip with a lot of space in the middle so I wrote a little program to convert a bitmap into metal and dropped a giant picture of the Stanford seal in the middle. But then nobody wanted to talk to me about the silicon cochlea design, they just wanted to know how I did the fancy logo. . .

1

@Andrew Wingate is it possible to grab an individual snapshot of our BISCUT die? Very much want to share with the team

Andrew Wingate

@Tholin more stills. Seems different things refract at different angles. But doesn't seem like discord is honoring the pic numbering. When I get to your die, I'll snatch one out and get some when I can hold it better by itself.

discord cares about selection order; on PC make sure they are sorted in the same order as you wish them to be shown on discord, and select them.

7:09 p.m.

I'm a bit suspicious about the traces being black looking on my FC01 design (the one die that has pads but no pad ring at all). I wonder if someone with a multimeter or such may feel like probing some of the obvious continuity/dis-continuity on one of them to sooth the worries.....

J-Lo

@Andrew Wingate is it possible to grab an individual snapshot of our BISCUT die? Very much want to share with the team

@InfoSecDJ offers pretty nice high quality imaging of dies, great to put in investor decks.

2

J-Lo

@Andrew Wingate is it possible to grab an individual snapshot of our BISCUT die? Very much want to share with the team

I can't seem to find BISCUT Can you share the project code? https://github.com/wafer-space/ws-run2 Also all of the dies should be represented above.

GitHub - wafer-space/ws-run2: wafer.space GF180MCU Run 2

wafer.space GF180MCU Run 2. Contribute to wafer-space/ws-run2 development by creating an account on GitHub.

Andrew Wingate

I can't seem to find BISCUT Can you share the project code? https://github.com/wafer-space/ws-run2 Also all of the dies should be represented above.

I think it’s COLF does that make sense?

J-Lo

I think it’s COLF does that make sense?

It's up there, this should be it https://cdn.discordapp.com/attachments/1419841819138854993/1559494764154323005/0041.JPG?ex=6ad33cad&is=6ad1eb2d&hm=c65fa9d9c3e998b60d601742f1ea1d635cd9a4fc1375f05a15faa3b15a9be97e&

Andrew Wingate

It's up there, this should be it https://cdn.discordapp.com/attachments/1419841819138854993/1559494764154323005/0041.JPG?ex=6ad33cad&is=6ad1eb2d&hm=c65fa9d9c3e998b60d601742f1ea1d635cd9a4fc1375f05a15faa3b15a9be97e&

Oh wonderful thank you!!

1

Andrew Wingate

I can't seem to find BISCUT Can you share the project code? https://github.com/wafer-space/ws-run2 Also all of the dies should be represented above.

@J-Lo - Your biscut one is not open source right? It's one of those ? secret designs?

Tim 'mithro' Ansell

@J-Lo - Your biscut one is not open source right? It's one of those ? secret designs?

correct, I guess I just took pictures of them all without thinking

@J-Lo - I mean, they requested it in a public channel - so I think the onus is on them.

Tim 'mithro' Ansell

@J-Lo - I mean, they requested it in a public channel - so I think the onus is on them.

Let me know how you want me to play the others? Not sure what, if anything you can glean from these pics, they're rather low res.

@Andrew Wingate - I need to get the terms of service up that say we don't guarantee confidentially and submitting a design means we can publicly publish stuff around your design for marketing purposes.

Tim 'mithro' Ansell

@Andrew Wingate - I need to get the terms of service up that say we don't guarantee confidentially and submitting a design means we can publicly publish stuff around your design for marketing purposes.

You want me to find the ones now and take them down tho? Looks like 11?

@Andrew Wingate - Also needed for the fact that people can purchase whole wafers and such.

Andrew Wingate

You want me to find the ones now and take them down tho? Looks like 11?

Only if people complain.

1

namibj

I'm a bit suspicious about the traces being black looking on my FC01 design (the one die that has pads but no pad ring at all). I wonder if someone with a multimeter or such may feel like probing some of the obvious continuity/dis-continuity on one of them to sooth the worries.....

Take this for what it is, but I rubbed my multimeter over the top of your die and saw signs of continuity, so it looks promising. I will try some better things in the future, but wanted to let you know.

9:23 p.m.

Also the blackness of the traces could just be an artifact of the way I lit them, they had some pretty harsh incident lighting as that was the best contrast.

@Tim 'mithro' Ansell @Andrew Wingate I’m not so concerned about a photo of the die, appreciate you thinking about it though. The RTL however…

J-Lo

@Tim 'mithro' Ansell @Andrew Wingate I’m not so concerned about a photo of the die, appreciate you thinking about it though. The RTL however…

RTL will never be shared. Everyone's always anxious and want to see die shots, so I tried to just get things out. Thanks for letting us know.

10:32 p.m.

@Tim 'mithro' Ansell Just riffing here, but in the future if we pass private: bool in the manifest or something, I can just have the machine take some pics and potentially stitch together a reticle and auto-blackout the ones we don't want *keep in mind the resolution leaves a lot to be desired. Also I'm unlikely to have this microscope in Sg, so that's another thought if we even want to keep doing this (I think we do)

J-Lo

@Tim 'mithro' Ansell @Andrew Wingate I’m not so concerned about a photo of the die, appreciate you thinking about it though. The RTL however…

I don't have access to your RTL, so can't accidentally share that

2

@Andrew Wingate May you please make a clear shot of RLCM too

wrong one, sorry.

Ghaith Al Sabagh

@Andrew Wingate May you please make a clear shot of RLCM too

ugh, I did have it right. They're all there https://cdn.discordapp.com/attachments/1419841819138854993/1559495248806150204/0053.JPG?ex=6ad33d21&is=6ad1eba1&hm=825c5b9f9f76b001d9350ee0974f170b57208b4a7470650913481720856c1bb5&****

Wow

Thx

1

Also, very pretty

Hope the inductors to be electrically pretty

3

Andrew Wingate

Take this for what it is, but I rubbed my multimeter over the top of your die and saw signs of continuity, so it looks promising. I will try some better things in the future, but wanted to let you know.

Idk why but I imagined you rubbing the entire meter over the wafer and divining the continuity

1

@Ghaith Al Sabagh - Are you behind https://github.com/EngGhaith/RLCM-Test-Passives-For-GF180MCUD ?

GitHub - EngGhaith/RLCM-Test-Passives-For-GF180MCUD

Contribute to EngGhaith/RLCM-Test-Passives-For-GF180MCUD development by creating an account on GitHub.

@Ghaith Al Sabagh

If so I have two pull requests for you;

https://github.com/EngGhaith/RLCM-Test-Passives-For-GF180MCUD/pull/1

https://github.com/EngGhaith/RLCM-Test-Passives-For-GF180MCUD/pull/2

Add CERN-OHL-P v2 license by mithro · Pull Request #1 · EngGhaith...

The wafer.space project page for RLCM (manufacturing ID G802RLCM) lists the design as licensed under the CERN Open Hardware Licence Version 2 - Permissive. Add the unmodified licence text published...

Add README and documentation of the test structures by mithro · Pu...

This adds documentation for the RLCM test chip on the wafer.space GF180MCU Run 2 shuttle (manufacturing ID G802RLCM). It is written from the layout files in the repository; nothing in the layout or...

@Ghaith Al Sabagh - I believe @baragonaru is also interested in helping test your chip

@lromor -^

Awesome! Thanks

Tim 'mithro' Ansell

@Ghaith Al Sabagh

If so I have two pull requests for you;

https://github.com/EngGhaith/RLCM-Test-Passives-For-GF180MCUD/pull/1

https://github.com/EngGhaith/RLCM-Test-Passives-For-GF180MCUD/pull/2

Thank you very much @Tim 'mithro' Ansell I would see them

Tim 'mithro' Ansell

@Ghaith Al Sabagh - I believe @baragonaru is also interested in helping test your chip

It would be also great

Ghaith Al Sabagh

Thank you very much @Tim 'mithro' Ansell I would see them

Accepted I would change some details later on after I test the dies. Thank you very much

Everyone should love logs me: Machine, you are being dumb, it's written in the logs machine: I do what you program me to do me: ... https://www.youtube.com/watch?v=GldgZFE4u-Y

Banana Neutron

The Log Song The Ren & Stimpy Show

1

12:08 a.m.

also ~1500 dies sorted today. Worked through some issues, tomorrow should be much better. Not sure how much more I want to go tonight.

1

Andrew Wingate

Take this for what it is, but I rubbed my multimeter over the top of your die and saw signs of continuity, so it looks promising. I will try some better things in the future, but wanted to let you know.

Yay! If you manage to tell that with e.g. the big square pads those that should be connected are but those that shouldn't aren't, that'd be all the hand held multimeter testing that's reasonably needed.

namibj

Yay! If you manage to tell that with e.g. the big square pads those that should be connected are but those that shouldn't aren't, that'd be all the hand held multimeter testing that's reasonably needed.

I'm not so sure anymore, I tried again under my microscope. My probe tips are gigantic compared to the pads. I wasn't able to get continuity anymore and any time I thought I did my probes were touching

‍♂️ I'll try again tomorrow with some better tools.

Andrew Wingate

also ~1500 dies sorted today. Worked through some issues, tomorrow should be much better. Not sure how much more I want to go tonight.

That's about one wafer worth right ?

1

Hi @Andrew Wingate Do you have any plan when the #2 run silicon would be shipped?

2

Ghaith Al Sabagh

Hi @Andrew Wingate Do you have any plan when the #2 run silicon would be shipped?

If you're getting bare dies, we will be shipping Tues Oct 27. The next day the rest are headed to China for packaging. I do not have a definitive date for that.

1

So this time it's dies on reels ? That's working good now ?

Yup and yup

3

Nice, congrats btw !

1

I was hoping to get my COBs before November 20th, or its gonna clash with some stuff I got happening around then.

3:33 p.m.

current cycle time is longer than it needs to be because of the feeders. You can see them pop a bit on the index, some were popping out at faster speed Before we sort again, I will be making some changes.

What's circular the trace left on the die ? Is it cracking them or just some glue ?

tnt

What's circular the trace left on the die ? Is it cracking them or just some glue ?

It's the picture

Oooh ... yeah now that you say it, I can sort of see

My bad ...

1

@Ghaith Al Sabagh beautiful chip in the feeder cam

1

Andrew Wingate

If you're getting bare dies, we will be shipping Tues Oct 27. The next day the rest are headed to China for packaging. I do not have a definitive date for that.

I'd like a majority (say, about 100?) of "my" FC01 dies with the tester and the few FlexMem COBs. Do we have an estimated shipping date for the COBs, assuming that things go as smoothly as can reasonably be expected? Ofc. no promises. More trying to estimate how soon I gotta be ready to probe them, such that I'm not to blame for delayed test results.

namibj

I'd like a majority (say, about 100?) of "my" FC01 dies with the tester and the few FlexMem COBs. Do we have an estimated shipping date for the COBs, assuming that things go as smoothly as can reasonably be expected? Ofc. no promises. More trying to estimate how soon I gotta be ready to probe them, such that I'm not to blame for delayed test results.

I have you marked as shipping from Singapore, they should be at fedex facility by or before Nov 2

Andrew Wingate

I have you marked as shipping from Singapore, they should be at fedex facility by or before Nov 2

the COBs happen in singapore?

No, those are marked as bare die?

4:54 a.m.

Not sure what the expectation is here, sorry, think I'm out of the loop on this one.

oh I just figured it'd be easier if they just come with the couple COBs and the tester

5:17 a.m.

i'm not gonna stop you from sending direct though

^^ @Tim 'mithro' Ansell

@namibj - Bare die shipments happen from Singapore, CoB ship from China

ohhhh

By no means anywhere near @BreakingTaps @psychogenic quality, beauty, value, etc. But made a short vid talking about the robot while it was running. Publish early, publish often Someday I will make something actually worth making public, but wanted to give you all a window into what we're doing. https://www.youtube.com/watch?v=kV5xYAFfbCw

Evezor

wafer.space run2 sorting robot

2

1

Andrew Wingate

By no means anywhere near @BreakingTaps @psychogenic quality, beauty, value, etc. But made a short vid talking about the robot while it was running. Publish early, publish often Someday I will make something actually worth making public, but wanted to give you all a window into what we're doing. https://www.youtube.com/watch?v=kV5xYAFfbCw

Interesting video and great to see how it's evolved/evolving!

2